针对传统方法在齿轮故障诊断中可靠性不高的问题,提出了基于证据理论的混合诊断算法。根据齿轮故障特征向量,采用两个并行的BP神经网络进行局部故障诊断,获得彼此独立的证据。再用

证据理论对各证据进行融合,最终实现对齿轮的故障诊断。实例结果表明,该方法可充分利用各种故障的冗余和互补信息,有效地提高诊断的可信度。

-For the reason of low-reliability exists in the gear fault diagnosis of traditional methods, a method

based on evidence theory hybrid diagnosis algorithm is presented. According to the fault feature vectors, two parallel

BP neural networks are used to carry on local fault diagnosis to acquire the independent evidences each other. Then

evidence theory is employed to fuse evidences, and gear fault diagnosis is fulfilled finally. Example shows that, vari-

ous faults redundant and complement information can be sufficiently used and the reliability of diagnosis is effectively

improved.